Awesome

Anterior Chamber Cell Detector (ACCDor)

This repository contains the code for ACCDor, an anterior chamber cell detector.

Installation

Create a new conda environment with Python 3.9:

conda create -n accdor python=3.9

Activate the conda environment:

conda activate accdor # sometimes could be `source activate accdor`

Install the required dependencies:

pip install -r requirements.txt

Checkpoints

ACCDor requires a pre-trained ViT model. Download the model from the ViT-H SAM model link. The link is provided in the SAM repository. Put the pre-trained model into the models folder.

You can also use

wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

to download the ViT model.

Getting Started

To process an image, segment the AC area, and detect cells, run the following command:

python -m apps.detect_cell

By default, this command will process the image located at data/example/example1.jpeg. The output will be saved in the data/output/{image_name} directory, where {image_name} is the name of the example image (in this case, example1).

Showcase

After running apps.detect_cell, intermediate stage images will also be generated. Below are examples of the generated images for the sample input image.

Original image

Anterior Chamber Mask

Cell Mask (Processed by Adjusted Cutoff)

Cell Mask (Discard the False Positive Candidate Cells)

Cell Dection by ACCDor

Citation

arXiv: https://arxiv.org/abs/2406.17577

To cite ACCDor in publications, please use:

@article{chen2024advancing,
      title={Advancing Cell Detection in Anterior Segment Optical Coherence Tomography Images}, 
      author={Boyu Chen and Ameenat L. Solebo and Paul Taylor},
      year={2024},
      journal={arXiv preprint arXiv:2406.17577}
}

Acknowledgements

Thanks to the support of AWS Doctoral Scholarship in Digital Innovation, awarded through the UCL Centre for Digital Innovation. We thank them for their generous support.